GIGO, Garbage in, Garbage out: An Urban Garbage Classification Dataset
Full paper can be found here:
Sukel, M., Rudinac, S., Worring, M. (2023). GIGO, Garbage In, Garbage Out: An Urban Garbage Classification Dataset. In: , et al. MultiMedia Modeling. MMM 2023. Lecture Notes in Computer Science, vol 13833. Springer, Cham. https://doi.org/10.1007/978-3-031-27077-2_41
This paper presents a real-world domain-specific dataset, which facilitates algorithm development and benchmarking on the challenging problem of multimodal classification of urban waste in street-level imagery.
The dataset, which we have named ``GIGO: Garbage in, Garbage out,'' consists of 24.999 images collected over a large geographic area of Amsterdam.
The capturing and annotating of the dataset took more than a year as part of a larger project investigating the potential of sensors for more sustainable and efficient waste collection. Our work aims at helping the cities with increasing populations, and thus more waste on the streets to collect different raw materials in a more sustainable fashion.
The collected data differs from existing benchmarking datasets, introducing unique scientific challenges. In this fine-grained classification dataset, the garbage categories are visually heterogeneous with different sizes, origins, materials, and visual appearance of the objects of interest.
Even though challenging, there is an abundance of urban data available in the geographical area of the collected data. Examples are information about demographics, different neighborhood statistics and information about buildings in the vicinity. This allows for experimentation with multimodal approaches.
Relationships within the dataset, information from demographics of the area, different neighborhood statistics, and information about buildings in the vicinity allows for different approaches to help solve the challenging task.
In addition, we provide several state-of-the-art baselines utilizing the different modalities of the dataset. Furthermore, we give suggestions on what can be done on the dataset, such as transformers to use the images' metadata effectively or graph structures that can process information between several images.
Additional contextual data can be found on maps.amsterdam.nl and data.amsterdam.nl.
